Operating system - CPU scheduling

• basic concept
  • Process execution is composed of CPU intervals and I/O intervals. The I/O main program has more short CPU intervals, and the CPU main program has more long CPU intervals.
  • CPU scheduler: preemptive scheduling, non-preemptive scheduling
  • Dispatch program: switch context, switch to user mode, jump to appropriate location in user program, restart program
    • Dispatch Latency: Dispatcher stops the time it takes for one process to start another
• Scheduling criteria: CPU usage, throughput, turnaround time, wait time, response time
• Scheduling Algorithm
  • First come first served FIFO
    • Escort effect: all other small processes wait for the big process to release the CPU
  • Shortest job priority scheduling: select the CPU with the shortest interval length
    • non-preemptive

      

    • Preemptive

      

  • Priority: preemptive, non-preemptive

    

    • starvation/infinite blocking: low priority cannot run
      • Solution: Aging, increase process priority with waiting time length
  • Rotation method: preemptive, the time slice is too large, the same as FIFO, and the context switching burden is too small, it should be greater than 80% of the CPU interval
  • multi-level queue
    • In-queue scheduling mechanism
    • Scheduling mechanism between queues
      • Fixed Priority Preemptive Scheduling
      • Divide time slices between queues
  • Multi-level feedback queue: processes can move between multiple queues, the implementation of aging
• multiprocessor scheduling
  • Symmetric multiprocessing: does not allow movement between multiple processors, each processor has its own scheduling algorithm
  • Asymmetric multiprocessing: allows movement between processors, only one processor schedules tasks
    • Process Soft Affinity: Allows movement between processors
    • Process Hard Affinity: Movement between processors is not allowed
  • Load Balancing: Keep all processors load balanced in SMP systems
• Thread scheduling: The scheduling object of a multi-threaded system is a thread
  • Local scheduling: Thread competition within the same process
  • Global scheduling: all system threads
• real-time scheduling
  • Hard real-time systems: strict deadlines
  • Soft real-time systems: no strict deadlines